Hanoi University of Science and Technology (HUST) stated on the morning of March 25 that it had received numerous inquiries regarding the scoring method for its TSA thinking assessment exam, following the announcement of the second session's results yesterday.

Specifically, the score distribution for the second TSA session, involving over 20,000 candidates, followed a standard distribution with an average score of 55,44 out of 100. Nguyen Tuan Dat, a student from Vinh Bao High School in Hai Phong, was the top scorer with 98,98 points. This marks the highest score in the exam's 6-year history.

Dat correctly answered 40 out of 40 questions in the mathematical thinking section, 18 out of 20 in reading comprehension, and 33 out of 40 in science/problem-solving.

On various forums, many questioned the scoring method, puzzled as to why the male student only had 1,02 points deducted despite answering 9 out of 100 questions incorrectly. Furthermore, the top scorer of the first TSA session in January answered more questions correctly (92 out of 100) but had a total score 2,88 points lower than Dat.

Associate Professor Dr. Vu Duy Hai, Head of Admissions and Career Guidance, explained that the TSA exam does not use the raw score calculation method prevalent in most Vietnamese exams today.

Elaborating further, he explained that typical exams calculate the total points from the questions on the test. For instance, if a test has 100 questions, each worth one point, a candidate answering 70 questions correctly would receive 70 points.

However, HUST does not use this method because it struggles to differentiate the abilities of candidates with the same raw score (e.g., both scoring 70 points) and fails to accurately assess their capabilities when exam difficulty varies.

Globally, to address this issue, large-scale exams such as SAT and ACT utilize modern measurement theories to estimate candidates' abilities accurately and reliably. One such theory is item response theory (IRT).

This theory posits that each candidate answering a question on an exam possesses a certain level of ability. A candidate with higher ability is more likely to answer any given question correctly than one with lower ability. Based on this, the university can quantify parameters related to question difficulty and discrimination, as well as candidate ability.

Hanoi University of Science and Technology applies the two-parameter IRT (difficulty and discrimination) model to the TSA exam. Under this model, a question is considered difficult if fewer people answer it correctly. Conversely, a question answered correctly by more candidates is deemed easier.

"With this model, the more difficult questions a candidate answers correctly, the higher their TSA score will be. Conversely, if a candidate only answers easy questions, their TSA score will be lower," Dr. Hai stated. "This explains why candidates can answer the same number of questions correctly but receive different TSA scores."

To ensure accuracy and fairness across different exam sessions, Dr. Hai noted that the university also applies a multi-stage scoring algorithm, processing common exam data from previous sessions to estimate parameters before calculating individual TSA scores for each candidate.

The four steps for two-parameter IRT scoring in TSA are as follows:

Step 1: Based on the performance of all candidates in each exam session, the scoring software estimates the difficulty and discrimination of each question in the current exam paper.

Step 2: The scoring software estimates each candidate's thinking ability score based on their performance and the current exam's scale.

Step 3: The current exam's scale is converted to a common reference scale used in previous exam sessions, bringing all candidate thinking ability scores to a unified metric.

Step 4: The referenced thinking ability score of the candidate is converted to the TSA score scale.

Thus, with the two-parameter IRT scoring process described above, the TSA score is not solely determined by the number of correct answers, but by the candidate's overall performance on the entire exam.

"This clarifies why the top scorer of the first session answered one more question correctly than the second session's top scorer, yet had a lower total score," Dr. Hai explained.

TSA thinking assessment exam score of the second session's top scorer, 2026. *Duong Tam*

Hanoi University of Science and Technology's TSA thinking assessment exam is one of the large-scale independent exams organized by the university for admission. This year, there is one more session scheduled for 16-17/5.

TSA thinking assessment exam 2026 registration and examination schedule

Last year, the average score for nearly 28,700 candidates taking the thinking assessment exam was 55,23 out of 100. Approximately 50 universities utilized these results for admissions. The benchmark scores for Hanoi University of Science and Technology in 2025, based on the thinking assessment exam, ranged from 46,48 to 86,97, with the data science and artificial intelligence program requiring the highest score.

Duong Tam